Over the course of 2 semesters (8 months) myself and 7 other students were tasked to design an autonomous drone capable of detecting and avoiding obstacles while flying in a GPS denied environment. This senior design project was sponsored by Lockheed Martin who allotted our team $1100 with an additional $550 for prototyping. The drone uses an NVIDIA Jetson Nano for object detection, a Pixhawk 4 flight controller, and a PX4flow optical flow sensor for localization. Because YOLO variants were prohibited for this project we used SSDlite Mobile Net V2 on the COCO dateset and used transfer learning to train our own dateset for hoop and pylon detection. Our flight code instructed the drone to takeoff, search for the closest obstacle, center itself, approach, and then perform a maneuver once a certain distance from the obstacle was reached. The drone was initially designed to be flown indoors but because of this pandemic we had to fly outside. This proved to be a huge issue with our initial plan to use the Intel Realsense T265 on the front because it didn’t perform too well outside. Overall, the project was still a major success.
thankyou my issue has been solved